Fix high memory consumption and code quality improvements

[georgeyiasemis]

What's changed:

• MRI models engines merged to MRIModelEngine. Each model_engine implements _do_iteration method.
• Fixes to load off memory during validation
• Main engine predict and evaluate methods refactored
• Disables stochasticity - set_all_seeds function refactored
• Memory leak identified and fixed (h5py 3.6.0 -> 3.3.0)
• Implement tests for the train method of Engine
• Black 22.1.0 fixes

Co-authored:

• @jonasteuwen

[codecov[bot]]

Codecov Report

Merging #172 (8968c55) into main (f624a66) will increase coverage by 1.31%. The diff coverage is 89.53%.

@@            Coverage Diff             @@
##             main     #172      +/-   ##
==========================================
+ Coverage   74.32%   75.63%   +1.31%     
==========================================
  Files          70       71       +1     
  Lines        6301     5093    -1208     
==========================================
- Hits         4683     3852     -831     
+ Misses       1618     1241     -377     

Impacted Files	Coverage Δ
direct/cli/train.py	0.00% <0.00%> (ø)
direct/cli/upload.py	0.00% <ø> (ø)
direct/data/bbox.py	71.42% <ø> (+17.14%)	:arrow_up:
direct/data/fake.py	97.84% <ø> (ø)
direct/environment.py	18.81% <0.00%> (+0.10%)	:arrow_up:
direct/functionals/__init__.py	100.00% <ø> (ø)
direct/functionals/challenges.py	83.67% <ø> (ø)
direct/inference.py	0.00% <ø> (ø)
direct/launch.py	0.00% <ø> (ø)
direct/nn/crossdomain/crossdomain.py	90.62% <0.00%> (ø)
... and 71 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update f624a66...8968c55. Read the comment docs.